A new evidence-based optimal control in healthcare delivery: A better clinical treatment management for septic patients

摘要

Treatment strategy of a realistic health care system must consider both system and measurement errors. The traditional optimal control method is commonly applied to deterministic systems instead of dynamic systems with uncertain errors. Therefore, this paper considers uncertain errors and stochastic characteristics in a dynamic health care system and proposes a new evidence-based optimal control (EBOC) approach that combines the traditional optimal control and machine learning methods. Four machine learning algorithms were tested, and the most suitable algorithm was combined with the traditional optimal control method for the sepsis model. Extensive computational studies proved that, compared to the traditional optimal control method, the EBOC method more efficiently controls disease progression and decreases total cost when uncertainty or measurement errors exist in the model, no matter the machine learning algorithm utilized. Moreover, the total c(n) settings are possible when numerous parameter combinations could affect control results, meaning determination of the optimal parameter set(s) becomes an NP-hardness problem. This paper also uses the genetic algorithm to find superior parameter settings to improve the performance and effectiveness of the control strategy created by the EBOC method.